From Pipeline Automation to Scientific Discovery: A Multi-Layer Agent-Driven Bioinformatics Framework

Abstract AI-driven bioinformatics is undergoing a multi-layer evolution from script automation to autonomous scientific discovery. We present a three-layer agent-driven framework spanning this complete spectrum. Layer 1 (Foundation Validation) benchmarks three OpenClaw agents—DESeq2 differential expression, GO/KEGG enrichment, and variant calling—against standard pipelines on TCGA liver cancer and E. coli genomic data, achieving near-perfect padj correlation (r ~ 1.0), >90% pathway overlap, and 95.8% variant concordance. Layer 2 (Domain Tool) presents CRISPR Agent, a multi-layer pipeline integrating Cas-OFFinder candidate discovery, multi-backend off-target scoring (mm_only/CFD/MIT), genomic context annotation, and end-to-end report synthesis, validated on GUIDE-seq/CIRCLE-seq/CHANGE-seq benchmarks. Layer 3 (Scientific Discovery) introduces SciOS, a ten-agent autonomous discovery system governed by a five-gate FORUM review chain (Plan-Execute-Audit-Synthesize-Ratify). Deployed across three single-cell and spatial transcriptomic datasets (457,000 cells, 154,000 spatial spots), SciOS produced three findings: (i) effector genes encoding ion channels, synaptic adhesion, and axon guidance proteins are significantly more conserved across cortical regions than transcription factor identity genes (P = 1.2 x 10^-16, d = 1.27)—a genuinely novel discovery inverting the standard neuroscience hierarchy; (ii) APOE expression is spatially coupled to oligodendrocyte progenitor cells (r = 0.673) but not mature oligodendrocytes (r = 0.136) in the human entorhinal cortex; and (iii) independent pseudobulk validation of CALB1/SOX6 dopamine neuron depletion in Parkinson's disease. The framework's Critic agent mandated average confidence reductions of -0.06 across all claims, while a Literature agent corrected a directional error introduced by autonomous analysis without domain context. These three layers collectively define an evolution from "doing" to "designing" to "discovering," establishing a methodological foundation for fully autonomous scientific discovery systems.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-17
DOI
https://doi.org/10.21203/rs.3.rs-11733080/v1
Primary Topic
Scientific Computing and Data Management
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

From Pipeline Automation to Scientific Discovery: A Multi-Layer Agent-Driven Bioinformatics Framework

Yuhan Zhang, Yujie Wang, Jue Huang
Research Square
Scientific Computing and Data Management
preprint

From Pipeline Automation to Scientific Discovery: A Multi-Layer Agent-Driven Bioinformatics Framework

Yuhan Zhang, Yujie Wang, Jue Huang
preprint en

Abstract

Abstract AI-driven bioinformatics is undergoing a multi-layer evolution from script automation to autonomous scientific discovery. We present a three-layer agent-driven framework spanning this complete spectrum. Layer 1 (Foundation Validation) benchmarks three OpenClaw agents—DESeq2 differential expression, GO/KEGG enrichment, and variant calling—against standard pipelines on TCGA liver cancer and E. coli genomic data, achieving near-perfect padj correlation (r ~ 1.0), >90% pathway overlap, and 95.8% variant concordance. Layer 2 (Domain Tool) presents CRISPR Agent, a multi-layer pipeline integrating Cas-OFFinder candidate discovery, multi-backend off-target scoring (mm_only/CFD/MIT), genomic context annotation, and end-to-end report synthesis, validated on GUIDE-seq/CIRCLE-seq/CHANGE-seq benchmarks. Layer 3 (Scientific Discovery) introduces SciOS, a ten-agent autonomous discovery system governed by a five-gate FORUM review chain (Plan-Execute-Audit-Synthesize-Ratify). Deployed across three single-cell and spatial transcriptomic datasets (457,000 cells, 154,000 spatial spots), SciOS produced three findings: (i) effector genes encoding ion channels, synaptic adhesion, and axon guidance proteins are significantly more conserved across cortical regions than transcription factor identity genes (P = 1.2 x 10^-16, d = 1.27)—a genuinely novel discovery inverting the standard neuroscience hierarchy; (ii) APOE expression is spatially coupled to oligodendrocyte progenitor cells (r = 0.673) but not mature oligodendrocytes (r = 0.136) in the human entorhinal cortex; and (iii) independent pseudobulk validation of CALB1/SOX6 dopamine neuron depletion in Parkinson's disease. The framework's Critic agent mandated average confidence reductions of -0.06 across all claims, while a Literature agent corrected a directional error introduced by autonomous analysis without domain context. These three layers collectively define an evolution from "doing" to "designing" to "discovering," establishing a methodological foundation for fully autonomous scientific discovery systems.

Research Square
Peking University (CN), Westlake University (CN)
Scientific Computing and Data Management
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon